Association between borderline dysnatremia and mortality insight into a new data mining approach

Table 4 Association between Dysnatremia and In-Hospital mortality. Comparison of the different regression models

Association Between Borderline Hyponatremia and Mortality
	OR (IC95%)	P	AIC
Classical Model^a	1.98 (1.73–2.68)	<.001	8098.4
PheWAS Model^b	2.59 (2.28–2.94)	<.001	11,002
Final Model^c	1.57 (1.35–1.81)	<.001	7585.5
Association Between Borderline Hypernatremia and Mortality
	OR (IC95%)	P	AIC
Classical Model^d	3.72(1.53–8.45)	<.001	6077.9
PheWAS Model^e	6.23(4.60–8.33)	<.001	8761.9
Final Model^f	3.47(2.43–4.90)	<.001	5920.8

Confounding Factors Retained in the different models:
^aClassical model: age, duration of hospital stay, number of ICD-10 codes, hospital admissions via the emergency department, ICU stay, dialysis, palliative care, Charlson Comorbidity Index
^bPheWAS model: A41, I20, I25, I48, I71, J15, J80, J96, K65, R07, R57, Z48, Z51
^cFinal model: classical model + I20, I25, I48, J80, R57, Z48, Z51
^dClassical model: age, duration of hospital stay, number of ICD-10 codes, hospital admissions via the emergency department, ICU stay, dialysis, palliative care, Charlson Comorbidity Index
^ePheWAS model: J69, J80, J96, N17, R57, S06
^fFinal model: classical model + J69, J80, R57, S06

ISSN: 1472-6947